I don't recognize that 🗣 🙉

Saturday, December 22, 2018 :: Tagged under: engineering. ⏰ 5 minutes.

Hey! Thanks for reading! Just a reminder that I wrote this some years ago, and may have much more complicated feelings about this topic than I did when I wrote it. Happy to elaborate, feel free to reach out to me! 😄

🎵 The song for this post is Drive Away, by Thomas Newman. 🎵

My partner Karen also writes software, and with another friend we worked together on a screaming chess set (a player moves a piece, and a speaker would play a sound of the piece begging for its life). We butted heads on how to model the chess logic.

I was modeling it like a functional programmer (all examples in Python):

# Create sum types with enums
Kind = Enum('Kind', 'Pawn Rook Knight Bishop Queen King')
Color = Enum('Color', 'Black White')

# Create product types with Namedtuples
Position = namedtuple('Position', [('x', int), ('y', int)])

Piece = namedtuple(
  [('kind', Kind), ('position', Position), ('color', Color)]

# Use toplevel pure functions to handle the logic + actions. This
# moves a piece. For want of a enumable `case` statement…
def move(piece : Piece, board : Board) ->  Piece:
    if piece.kind is Kind.Knight:
        # Move logic for a knight, returns a new piece with
        # altered position

    elsif piece.kind is Kind.Bishop:
        # Move logic for a bishop...

def can_capture(piece : Piece, board : Board) -> List[Piece]:
    if piece.kind is Kind.Knight:
        # Look at what a Knight can capture…

    elsif piece.kind is Kind.Bishop:
        # Look at what a Bishop can capture…

Karen was modelling it in a more classical OOP fashion:

# Define a class that others can inherit from
class Piece(object):
    # Common members are stored in the base class, available to subclasses.
    def __init__(position, color):
        self.position = position
        self.color = color

    def move(self):

    def can_capture(self, board):

# Subclasses, specialize on behavior
class Knight(Piece):
    def move(self):
        # Move logic for a Knight, mutates the instance

    def can_capture(self, board):
        # Look at what a Knight can capture given a board

class Bishop(Piece):
    def move(self):
        # Move logic for a Bishop...

    def can_capture(self, board):
        # Look at what a Bishop can capture, given a board

Karen's reaction to my design was somewhat aghast. "You're losing the encapsulation that OOP provides! Now your code related to Knights is sprinkled in 10 different functions! If you want to learn or change something about a Knight, you have to hunt around a ton of different functions!"

My reaction to Karen's design was somewhat aghast. "You're losing all the encapsulation that FP provides! Now all your code related to moving is in 10 different classes! If you want to learn or change something about movement, you have to hunt around a ton of different object definitions!"


So… we're both right. Karen's preference was to have richer objects that know how to operate on themselves, mine was to have immutable objects with toplevel functions that return new instances. Her approach doesn't let her have a single move definition, and mine doesn't let me have a single place to look for Knight-related functionality. Both would work perfectly fine.

Our reactions misunderstood each other's preferences and needs. We both wanted to solve the problem of "model chess" but for me it was more an issue of limiting any complexity brought about by mutable state, for Karen it was about consolidating the datatypes and their functionality, hiding their behavior from others, and finding opportunities for re-use.

So let's talk about CSS-In-JS 😄

To recap: CSS was introduced in 1996, got much more popular in the early aughts with delights like the CSS Zen Garden. It aims to fix "how should a page be rendered" with toplevel, global, declarative statements based on the structure of your page.

CSS-In-JS seems to be the opposite of this: CSS-In-JS takes the "componentization" of our pages to an extreme and lets you style granular, localized statements to only the component you're currently looking at.

Looking at this Twitter thread, it seems like most people with a horse in this race are talking past each other. The CSS-In-JS people dislike that the traditional stylesheet rules are global, that they cascade, and that they're lexically decoupled from the components they're working on, whereas those against CSS-In-JS dislike it precisely because it takes those properties away.


Stepping back, I personally prefer CSS-As-Designed than CSS-In-JS, but the emergence of CSS-In-JS doesn't surprise me. Over the last 20 years, our pages became apps, those apps have more users than the populations of most nations, and our Webmasters became teams of dozens or hundreds of full-time software developers working together to support every screen size for hundreds of millions of people who never log off.

CSS is, frankly, perfect if you view the web as a set of linked documents whose content is mostly textual, represented by persistent and meaningful URLs. That's exactly what this blog is, and after you get past the browser-normalizing bit, you get something rather readable and high-performance in very few lines.

I also believe its much easier to work with a stylesheet if you're a smaller team of people working together with little turnover. You can build and enforce conventions. You have institutional memory. You don't have too many other toes to step on.

Most Tastemakers™ in tech are extremely large companies who are flexing how they scale their processes to hundreds or thousands of engineers for products servicing hundreds of millions of users. Just because something doesn't scale to a larger, more glorious case, doesn't mean it's a bad choice! Facebook uses technologies like React and CSS-In-JS, but that doesn't mean you have to use them; neither does it mean you're wrong if you want to. Saying that CSS-As-Designed doesn't happily scale to Facebook's engineering teams shouldn't disqualify it from your project or your company.

But if you are Facebook (or about 10 other companies), then yes, you have so many fingers working on the pudding at the same time that it's probably a blessing to have your work gated off from someone else.

Most tech fighting is people saying true statements about a technology to argue some variation of "I don't yet recognize your what tradeoffs you're making, and I can't imagine why they're valid." These are often said with a ton of invective.

I wish we'd stop doing that! I also recognize it's hard (you can probably find a ton of it on this blog, and not even going that far back). Awareness is a good first step, and I think CSS-In-JS is a strong example to learn from.

(This topic is well-covered, with a different set of examples, in this talk.)

1. ^ SICP looks into almost the same tradeoff, though using a pretty different vocabulary. It also taps into the "closures are a poor man's objects; objects are a poor man's closures" refrain.

Thanks for the read! Disagreed? Violent agreement!? Feel free to join my mailing list, drop me a line at , or leave a comment below! I'd love to hear from you 😄